First SaaS: Summaries from Youtube Videos - background story

Hello everyone,
I am new in the indiehacker and SaaS community. I'm a freelance developer from germany. At the moment I have no proper client project going on and that gave me the chance to put my skills to test and try to create my first own product.

Personally I am pretty addicted to youtube. Thats the main media I consumer. Still my watch later list is always at capacity. Thats why I created https://summarizeit.io/ it's an AI powered service, that creates summaries from youtube videos.

example summary

At the moment I am struggling with the marketing part, I am not good at the "build a community first" approach, and just jumped straight into coding. Right now I am having about 50 visitors pay day, but no customer yet.

But the way I see it: Even if I won't get subscribers, I still learn a lot from this project and it's something interesting to put on my resume.

I want to share my way to the product with you:

💡 The prototype

Building the prototype was easy. It was just a couple lines of code of node.js to fetch a youtube transcript and send that to gpt-3 to summarize it.

🎨 The design

Design is not really my greatest skill, so I wanted to take midjourney for inspiration. I asked somethin like "Design for SaaS for VideoSummaries" and got this result:
first design from midjourney
I didn't really like the concept with the big background image, but I liked the color scheme, so I picked some of the colors for later use.

💻 The stack

I worked the last couple years with react, so I decided to stay with that.

Next.js
Supabase
Stripe
Netlify
OpenAi

I am using next.js api routes, it's pretty cool that netlify automatically turns them into netlify functions and those run on aws lambda. So I have a super hasslefree deploy.

💪 The Challenges

Long Videos.

gpt-3 only allows a maximum amount of tokens. so even videos of just medium length often exceeded this limited. To solve that I created a function, that basically runs the summary in a loop: split the whole text into chunks, summarize them, put together, see if it makes sense, repeat.

Serverless functions

As mentioned above all my stuff is in serverless functions from netlify. They have a runtime of max 10s. At the moment the open ai servers are always on heavy load and often even take longer than 10s to return something.

I decided to use background functions from netlify to solve that. They can run about 15minutes. Only drawback with them is, that they only return status 202, that they are invoked, but don't return a result to the client. So what I do is to invoke the function and then on supabase use the realtime feature, to listen for when the summary is added to database. Then the user gets a notification, that the summary is available now and if he is still in the loading screen he gets pushed to the newly created summary.

So one thing for the future could be, to set up my own server, instead of using netlify functions.

💰 The marketing

As I said, I didn't take the route to do marketing before launch. I'm just building the tool and kind of launch as I go. What I did so far for marketing:

a couple of posts on reddit, mostly in SaaS groups
instagram, where I present motivational summaries from ted talks, gary vee, and so on https://www.instagram.com/summarizeit.io/
twitter where I post news about what I'm building at the moment https://twitter.com/home
all previously generated summaries are stored and free to see, I create a sitemap with all summary links and push that to google, I hope, that this will create interesting links for seo over time.

📈 The next steps

I have some streamer friends and will ask them to try SummarizeIt on stream. That way I can see how people interact with it and maybe some viewers want to try it out too.
I think the product is in a "good enough" state now, where I will leave it as is, and try to see if I can get subscribers. If yes, then I may continue on features and stability, if not I am still happy, that I learned a lot.

Big shoutout too indiehackers and the SaaS communities on Reddit for encouraging me to build a product myself. I've been thinking about that I should build something for many years now, but I always was like "naah, thats just wasted time".

Oh and for anyone wondering about how much work I put into it. I built the first prototype middle of december and since then I have about 140 hours of coding.

Feel free to try the software. Theres a free tier after you register. Even if you don't want to try it, I would appreciate feedback on the landing page.